This document discusses the scientific basis of Firstbeat Technologies' physiological analysis. It summarizes that Firstbeat has:
- Published over 90 peer-reviewed scientific papers validating its methods through laboratory and real-world testing.
- Collected data from thousands of laboratory tests and over 250,000 long-term heart rate variability monitoring sessions.
- Demonstrated accuracy in measuring variables like oxygen consumption, energy expenditure, VO2 max, and stress/recovery through comparisons with gold-standard methods.
- Validated its stress analysis through studies manipulating the autonomic nervous system, using stress tests, and real-life observational studies.
JavaScript Usage Statistics 2024 - The Ultimate Guide
Firstbeat and the Science of Personalized Insight for Everyone - Dr. Ilkka Korhonen, VP of Technologies, Firstbeat
1.
2. Ilkka Korhonen, VP Technologies
HRV Summit 2018
London, UK
May 3, 2018
The Science of Personalized Insights for Everyone
3. FIRSTBEAT IS SCIENCE-BASED
• 90+ scientific peer-reviewed publications and
papers
• Firstbeat in-house research
• Thousands of tests in the laboratory
• >250.000 long-term HRV monitoring sessions
performed in the field
• >1.8M training sessions completed by athletes
• Firstbeat analysis brings near-laboratory accuracy to
the field and for everybody
• Oxygen consumption
• VO2Max
• Energy expenditure
• EPOC and training effect
• Stress and recovery
For more information, please see https://www.firstbeat.com/en/science-and-physiology/white-papers-and-publications/
4. FIRSTBEAT IS SCIENCE-BASED
• 90+ scientific peer-reviewed publications and
papers
• Firstbeat in-house research
• Thousands of tests in the laboratory
• >250.000 long-term HRV monitoring sessions
performed in the field
• >1.8M training sessions completed by athletes
• Firstbeat analysis brings near-laboratory accuracy to
the field and for everybody
• Oxygen consumption
• VO2Max
• Energy expenditure
• EPOC and training effect
• Stress and recovery
For more information, please see https://www.firstbeat.com/en/science-and-physiology/white-papers-and-publications/
9. HRV indexes, respiration frequency, VO2, energy expenditure, moving speed etc...∑
Does the segment contain signs of physical activity?
(VO2, %VO2max, movement…)
Is parasympathetic activity dominating?
Is sympathetic activity dominating?
YES
YES
Physical activity
Recovery
StressYES
No
No
Other stateNO
MEASUREMENT DATA:
Beat-by-beat HR (HRV)
3-D accelerometer
USER:
• Age
• Height
• Weight
• Gender
• Activity level
ADVANCED DIGITAL MODEL OF BODY’S REACTIONS
10. HRV indexes, respiration frequency, VO2, energy expenditure, moving speed etc...∑
Does the segment contain signs of physical activity?
(VO2, %VO2max, movement…)
Is parasympathetic activity dominating?
Is sympathetic activity dominating?
YES
YES
Physical activity
Recovery
StressYES
No
No
Other stateNO
MEASUREMENT DATA:
Beat-by-beat HR (HRV)
3-D accelerometer
USER:
• Age
• Height
• Weight
• Gender
• Activity level
ADVANCED DIGITAL MODEL OF BODY’S REACTIONS
11. WHY FIRSTBEAT VO2MAX IS ACCURATE
Good. Data
included.
HR
Running speed
Stop at traffic
lights
Under bridge, in a
tunnel (with GPS)
Poor heart rate
signal
Estimation
reliability:
Only good
data
included
Low. Data
excluded.
Low. Data
excluded.
Good. Data
included.
Low. Data
excluded.
Good. Data
included.
Good. Data
included.
Walk/run slowly Walk/run fast.Walk/run
steadily.
Walk/run
variable pace.
HR and
speed data
User action
Accuracy:
• Running or walking (speed with GPS): MAPE* ~5-6%
• Walking (speed with accelerometer based steps): MAPE* ~6-9%
• Cycling: MAPE* ~8-9%
Accounts for:
• Up and downhills
• Sudden starts and stops in
activity
• Different speeds, variable speed
• User initial VO2max based on
age, height, weight, gender
• HR drift
• Heat and altitude (coming soon)
Internal reliability measure:
• HR data quality
• Physiologically possible HR &
speed reactions
• Accounts unreliability due to e.g.,
changes in surface (soft/hard)
Output reliability weighted average
between measurements
12. VO2MAX END EE VALIDATION WITH OPTICAL HR DATA
• Firstbeat method was used with a wrist-worn optical heart rate
device to estimate energy expenditure (EE) and aerobic fitness
(VO2max).
• Firstbeat method used with optical data and submaximal self-paced
outdoor running results in closely similar estimation of VO2max
than the reference test in a sports laboratory (MAPE 5.2 %).
• Energy expenditure was more accurate during moderate to heavy
intensity activities (MAPE 6.7 %) than light intensity activities.
• OHR combined with Firstbeat physiological modeling may be
used for accurate estimation of VO2max and EE.
Parak et al. (2017). Estimating heart rate, energy expenditure, and
physical performance with a wrist photoplethysmographic device
during running. JMIR Mhealth and Uhealth 5(7): e97.
All (n=24)
Male
(n=13)
Female
(n=11)
Bias (ml*kg-1
*min-1
) -1.07 -1.28 -0.82
MAE3
(ml*kg-1
*min-1
) 2.39 2.29 2.51
MAPE4
5.2% 4.7% 5.8%
Statistical test p=0.059(W) p=0.082(T) p=0.416(T)
Correlation coefficient 0.86**(S) 0.77**(P) 0.69*(P)
OHR vs respiratory gas analysis based VO2Max
Activity All (n=23) Male (n=12) Female (n=11)
Bias (kcal) -5.58 -6.78 -4.28
MAE3
(kcal) 7.52 10.43 4.34
MAPE4
6.7% 8.2% 5.1%
Stat tests p<0.007**(T) p=0.081(T) p=0.001**(T)
Corr. coef. 0.97**(P) 0.93**(P) 0.99**(P)
OHR vs respiratory gas analysis based EE
14. ANS blocking
Manipulate physiology, block
ANS branch directly.
Stress test
Manipulate condition, subject
people to stressful or relaxing
condition
Real-life setting
Observe people during real-
life, measure physiological
stress & subjective stress
VALIDATION OF STRESSRRI,ms
0 1 2 3 4 5 6 7 8
400
500
600
700
800
900
1000
aika (min)
RR(ms)
pre
post 5
post 4
post 3
post 2
post 1
istumassa O seisomassa
RRI,ms
sitting standing
time, min
Expected max correspondence with HRV measured stress
High to very high Moderate to high Moderate to low
15. QUANTIFYING STRESS: ANS BLOCKADE STUDY
• Pre = HRV baseline prior the test
• Post1 = full blockade of ANS
parasympathetic (vagal) branch
• During Post2 – Post5 HRV is
regained as vagal blockade is
cleared from the body
0 1 2 3 4 5 6 7 8
400
500
600
700
800
900
1000
aika (min)
RR(ms)
pre
post 5
post 4
post 3
post 2
post 1
istumassa O seisomassa
RRI,ms
sitting standing
time, min
HR & HRV after ANS blockade
Ability of short-time Fourier transform method to detect transient changes in vagal effects on
hearts: a pharmacological blocking study. Am J Physiol Heart Circ Physiol 290: H2582–H2589,
2006. Kaisu Martinmäki, Heikki Rusko, Sami Saalasti, and Joni Kettunen.
16. TRIERS SOCIAL STRESS TEST (TSST) STRESS TEST
Rest1 10’ Speech preparation 10’ Public
speech 5’
Mental
arithmetic 5’
Rest2 10’
Purpose: to cause high stress levels to the subjects to validate stress detection
1. Baseline rest 10 min
2. Public speech
preparation 10 min
3. Public speech (in front
of 2 judges,
videotaped) 5 min
4. Mental arithmetic
(counting backwards
from number 1022 in
steps of 13) 5 min
5. Rest 10 min
17. STRESS TEST RESULTS
• Stress is detected as expected during stressful test phase, and recovery dominates during rest
periods
• ECG and PPG derived results are well aligned
0
10
20
30
40
50
60
70
80
90
100
Rest1 Preparation Speech Numbers Rest2
Relaxation %
EKG ETE PPG
0
10
20
30
40
50
60
70
80
90
100
Rest1 Preparation Speech Numbers Rest2
Stress %
EKG ETE PPG
Recovery Stress
18. SCIENTIFIC VALIDATION STUDIES: VO2, EE
Feature Findings / accuracy
Oxygen
consumption
- Strong correlation (r=0.98) between gold standard and Firstbeat’s oxygen consumption, SEE for
oxygen consumption was 2.6 ml/kg/min with %CV) of 6.0% during exercise (Montgomery et al.
2009)
- Strong correlation (r=0.96) between gold standard and Firstbeat EE, SEE of energy expenditure
was 6.14 kJ/min with %CV of 12.7% during exercise (Montgomery et al. 2009)
- Average deviation 0.6-1.0 METs (2.1-3.5 ml/kg/min) and coefficient of determination R2=0.77-
0.94 of Firstbeat VO2 compared to gold standard in real-life conditions (Smolander et al. 2011)
- Mean absolute error (MAE) 1.9 ml/kg/min in VO2 across all subjects and intensity ranges
(n=32) – an accuracy improvement of 48% to basic HR-based VO2 (Pulkkinen et al. 2004)
Energy
expenditure
- Strong correlation (r=0.97) in maximal exercise test and (r=0.72) in light walking between gold
standard EE and Firstbeat EE when using correct HRmax and VO2max (Robertson et al. 2015)
- Differences in mean EE only 0.4-5.7% during exercise test and only 0.6-6.1% during 30min walk
against gold standard (Robertson et al. 2015)
- Coefficient of determination R2=0.81 across all conditions from real-life tasks to maximal
exercise (Pulkkinen et al. 2004)
- MAPE was 6.7% in heavy activity against gold standard and 16.7% in light activity when using
optical HR device (submitted for publication by Parak et al. 2016)
19. Feature Findings / accuracy
VO2max
(Fitness level)
- The accuracy was 95% (MAPE ~5%) against gold standard in running (n=79, 2690 freely
performed runs, VO2max measured 4 times) (Firstbeat 2015)
- The accuracy was 92% (MAPE ~5%) against gold standard in cycling (n=29 cyclists) with power
and HR data (Firstbeat 2015)
- Device use case MAPE was 4.9% (several measurements from a single person) and 5.3% based
on only one measurement (Firstbeat database 2016)
- The accuracy was 96% (MAPE 4.3%) against gold standard in running on track when correct
HRmax was used (Brenner et al. 2011)
- MAPE was 5.2% and correlation coefficient (r=0.87) against gold standard in running when
using optical HR device (Parak et al. 2016)
EPOC &
Training Effect
- Strong correlation (r=0.89) between Firstbeat’s estimated EPOC and measured gold-standard
EPOC (Rusko et al. 2003)
- Tight connection (r2=0.79) between Firstbeat EPOC and blood lactate (Rusko et al. 2003)
Anaerobic
threshold
- Firstbeat reported accuracy: MAPE 1.9% for HR at threshold, i.e. ~3.5 bpm, MAPE 6.2% for
speed at threshold (Firstbeat database 2016)
SCIENTIFIC VALIDATION STUDIES: VO2Max, EPOC, TE, Anaerobic threshold
20. SCIENTIFIC VALIDATION STUDIES: STRESS AND RECOVERY
Feature Findings / accuracy
Stress and
recovery
- Subjective stress was directly associated with Firstbeat stress (P<0.05) and inversely with
Firstbeat recovery (P<0.05) in psychologically distressed persons (Föhr et al. 2015)
- Association (P<0.01) between self-reported burnout symptoms and Firstbeat stress (Teisala
et al. 2014)
- Association between daily emotions at work and chronic work-related stress with Firstbeat
recovery (P<0.01-0.05) (Uusitalo et al. 2011)
- Significant correlation between awakening cortisol response and stress (r=-0.49) and
relaxation during sleep (r=0.50) (Rusko et al. 2006)
- >30 other scientific studies where Firstbeat analysis has been successfully applied to analyse
stress and recovery during various research settings
21. • Firstbeat is science-based
• Models built based on physiological knowledge and optimized with
experimental data against golden references
• Laboratory – thousands of recordings
• Real life – hundreds of thousands of recordings
• >90 scientific studies showing validity and application published so far – and
growing
• When choosing your method: ask for evidence and assess that with critical
look!
CONCLUSIONS